Microsoft and Nvidia to invest up to $15bn combined in Anthropic; Anthropic commits $30bn to Azure

Microsoft and Nvidia announced they would together invest up to $15bn in Anthropic — Microsoft up to $10bn and Nvidia up to $5bn — in a deal structured around cloud compute rather than cash alone. In return, Anthropic committed to purchase $30bn of Azure compute capacity and to contract for up to a gigawatt of additional capacity built on Nvidia’s Grace Blackwell and Vera Rubin systems.

The arrangement had three parts: Anthropic scaling its Claude models on Microsoft’s Azure cloud, a technical collaboration between Anthropic and Nvidia on optimising model performance and efficiency for Nvidia’s hardware, and expanded distribution of Claude through Microsoft’s commercial products, including GitHub Copilot and Microsoft 365 Copilot. Azure customers gained access to Anthropic’s frontier models — Claude Sonnet 4.5, Claude Opus 4.1 and Claude Haiku 4.5 — making Claude available across all three major cloud platforms (Azure, AWS and Google Cloud) for the first time.

Anthropic’s announcement was careful to state that Amazon remained its “primary cloud provider and training partner,” a qualification that mattered because Amazon has invested heavily in Anthropic and hosts the bulk of its training infrastructure on Trainium chips. The Microsoft-Nvidia deal did not replace that relationship but added a second major compute channel alongside it.

The structure — a chipmaker and cloud provider investing capital that a lab is obliged to spend back on their own compute and hardware — extended a pattern already visible in Nvidia’s investment in OpenAI and Microsoft’s earlier commitments to the same company. Critics of the arrangement, including some industry analysts, argued that circular deals of this kind inflate reported revenue and investment figures without new money entering the system, since the capital returns to the investor as compute purchases; the companies did not address this framing directly in their announcement.